MVP build tracker — dependency-ordered plan

Single view of the build. Governing spec: docs/superpowers/specs/2026-08-22-farewright-mvp-design.md. Foundation plan: docs/superpowers/plans/2026-08-22-contract-core-foundation.md. Each issue lands through a linked PR with tests/evidence; review gates are hard stops.

Milestone 0 — Human accounts & M0-F spike (parallel track)

• #2 Provision external demo accounts (human-only — founder)
• #4 Throwaway remote MCP host/auth feasibility slice
• #5 Record observed auth, client-registration, monitoring, and browser-link rulings

Milestone 1 — Contracts & core foundation (dependency order)

• #3 Repository, CI, templates, branch/review enforcement
• #6 Apply merged-spec and blocking-audit rulings
• #7 Domain event/command/state/projection/money contract
• #8 Supplier/search-result/policy-evaluation contract
• #9 Threat model + external-effect failure matrix
• #10 GATE: independent complete M0-C review — waits for #6/#7/#8/#9 and relevant #5 rulings
• #11 Event log + real projectors/rebuild
• #12 Atomic command receipts/idempotency/effect planning
• #13 Transactional state enforcement + action/payment-level USD ledger

Milestones 2–5 (open after #10; issue dependencies still apply)

2 — Vault & effects: #15 waits for #11; #14 waits for #11/#12. They may proceed in parallel against reviewed redaction/effect interfaces.

3 — MCP surface, setup skill & supplier stubs: #16 waits for #11/#12/#15; #18 waits for #13/#14 → #17 setup skill after #16/#18

4 — Auth & payments: #19 OAuth/RBAC/revocation → #20 approval pages and #21 Stripe lifecycle → #22 GATE: security/effects review

5 — Service loop & consoles: #23 service loop → #24 property console; #25 observability/SSE → #26 redaction proof + claim integrity

Milestone 6 — Integration, deploy & demo readiness

• #27 Duffel adapter + effect reconciliation
• #28 Monitoring Path A only as proven in #5; Path B always available
• #29 Director: clock, disruption injection, labeled fixture fallback
• #30 Railway staging/demo profiles, health and secret checks
• #31 Live-test plus honest outage/fallback rehearsals
• #34 Private reliability matrix + phase-seed index (QA-only; never input steering)
• #32 Cold-user rehearsals with user-chosen inputs and no-match/refusal accepted
• #33 GATE: independent demo-integrity review

Standing rules

• The implementing and adversarial-review models are independent for each revision; roles are not assigned permanently by model name.
• GitHub issues/milestones/PRs/CI are the durable execution record; a Project board is optional.
• Secrets remain founder-only in provider dashboards or local secret storage, never issues/code/model context.
• The trip-operating human is called user in product-facing contexts and traveler in domain contexts; meeting-audience labels never become implementation identifiers or test personas.
• The agent interprets, ranks, recommends, and selects; Farewright stores the graph, evaluates confirmed typed hard rules, and revalidates/enforces them at execution.
• Search returns provider-completeness outcomes and candidate-local policy evidence, never authority or a Farewright recommendation.
• The user chooses the route, dates, USD budget, threshold, required constraints, soft preferences, and selections. Honest no-match, prohibition/refusal, and unavailable outcomes are valid.

Start with docs/superpowers/specs/2026-08-22-farewright-mvp-design.md and docs/superpowers/plans/2026-08-22-contract-core-foundation.md, then follow the numbered issue dependencies in this tracker. The work is complete when the listed milestones, linked PRs, tests or evidence, and review gates are resolved.